Matrix Multiply with Vector and Tensor in Python

Question

I have a Vector, M, with size N and a Tensor, d, with size NxNxD.

My aim is to perform the matrix multication M*d[i,:,:] for each i to get a new matrix with size nxD.

Now I could just do it like this:

a = np.zeros((n,D))
for i in range(n):
    a[i] = G*np.matmul(M,d[i,:,:])

but I'd prefer a one line solution if it exists. Anyone got any ideas?

you could try to write it as list comprehension - and code would be in one line. OR you can put it in function and later you can execute function in one line (and you can reuse it in other places). OR maybe it has function .apply() to execute function on every row or column. — furas
– furas, Commented Jul 13 at 11:33
you could create minimal working code with small example data. It could be used to create answer. And it may show if you use torch, tensorflow or other module for tensor. — furas
– furas, Commented Jul 13 at 12:17
What? No? numpy is perfectly capable of dealing with tensors, and you definitely shouldn't use a comprehension. — Reinderien
– Reinderien, Commented Jul 13 at 12:39
@Reinderien I expected that numpy may do it directly but I don't know numpy that well to do it :) So I suggested other methods. — furas
– furas, Commented Jul 13 at 13:15

Reinderien · Accepted Answer · 2025-07-13 12:35:45Z

4

Don't use a loop or a comprehension. Use einsum:

G = 0.5
d = rand.integers(low=0, high=10, size=(3, 3, 2))
M = rand.integers(low=0, high=10, size=3)
a = G*np.einsum('ijk,j->ik', d, M)
print(a)

answered Jul 13 at 12:35

Reinderien

16.8k9 gold badges57 silver badges93 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Actually perfect, thanks